Public health evaluation: which designs work, for whom and under what circumstances?

摘要

Properly designed and conducted randomised controlled trials (RCTs) represent the most internally valid means of estimating the effectiveness of complex interventions.~1 However, RCTs are often dismissed as being unsuitable for the evaluation of public health interventions, in a debate often dominated by arguments based on misunderstandings, misrepresentations and ideological objections. Macintyre~2 counters many of these arguments and argues that 'RCTs are both more possible than many objectors think, and more conclusive about the benefits and cost effectiveness of policies and interventions [than other designs]'. The paper by Bonell and colleagues (see page 582)~3 and its companion paper by Cousens and colleagues (see page 576)~4 provide welcome contributions to this debate, by seeking to delineate circumstances in which RCTs may not be possible and describing various analytical approaches to compensate for the lack of randomisation in order to derive unbiased estimates of effect. Bonell et al~3 identify three key features of RCTs, random allocation, control groups and prospective follow-up, of which only the first is unique to RCTs.